论文部分内容阅读
在细胞核仁银染(AgNORs)图象分析中关键的一步就是通过阈值分割将细胞核和银染颗粒提取出来,以准确地计算出细胞核仁和核的面积、核仁与核的面积之比(I.S％)、核仁与核的积分光密度比值(I.O.D％)等重要参数,为医生进行肿瘤的鉴别诊断提供量化数据。本文提出了一种新的阈值分割算法,采用(红R+绿G+蓝B)/3彩色图象特征作为图象的特征信息,对直方圆数据进行高斯平滑并根据爬山聚类法计算出直方图峰的近似位置、连接相邻两峰以构成一凸多边形并对该凸多边形进行凹性分析,计算出的凸残差最大处就是所要求的阈值T。通过对腺癌细胞和淋巴瘤细胞图象的实际计算,证明该方法是快速有效的。
A key step in the image analysis of AgNORs is the extraction of nuclei and silver-stained particles by threshold segmentation to accurately calculate the area of nucleoli and nuclei, the ratio of kernel area to nucleus (IS% ), And kernel integral optical density ratio (IOD%) and other important parameters for the diagnosis of the tumor to provide doctors with quantitative data. In this paper, a new threshold segmentation algorithm is proposed, which uses the features of (red R + green G + blue B) / 3 color image as the feature information of the image, performs Gaussian smoothing on the data of the histogram and calculates the histogram according to the mountain climbing clustering method The approximate location of the peak, connecting the adjacent two peaks to form a convex polygon and concavity analysis of the convex polygon, the calculated maximum convex residual is the required threshold T. Through the actual calculation of the images of adenocarcinoma cells and lymphoma cells, the method is proved to be fast and effective.